Inline elements such as SPAN, I, B, FONT, etc. aren't currently kept in the HTMLDocument tree. The relevant information is stored as the character content attributes. This results in a number of problems.

  * Element addressing. There's no way to locate inline HTML elements (by their IDs for example) in the parsed document tree as they don't actually exist. This makes it difficult to properly implement ID-based CSS locators or DOM API basics.

  * CSS box model. It's not possible to implement the basic CSS functionality like borders and backgrounds for the inline elements.

  * I/O fidelity. The HTML serializer output does not correspond to the input document read by the HTML parser. This confuses users.

The inline HTML elements should be preserved in the parsed document tree in a way that a) makes it possible to address the above problems and b) is compatible with the current element API.